Can a Z Score Be Negative?
Yes — and not as an edge case. Exactly half of every normal distribution has negative z scores, by construction. A negative z score simply means the value falls below the mean; the size of the number says how far below, in standard deviations.
Short answer: A z score is negative whenever the raw value X is smaller than the mean μ, because z = (X − μ) / σ and the numerator turns negative. z = −1.5 means "1.5 standard deviations below the mean" — the 6.68th percentile.
The one condition that produces a negative z score
The formula z = (X − μ) / σ has three moving parts, and only one of them can change the sign. The standard deviation σ is always positive (it is a spread, and spreads cannot be negative). The mean μ is just a location. So the sign of z is the sign of (X − μ): value below the mean, negative z; above, positive; exactly equal, zero. Textbook variants of the question — "what conditions would produce a negative z score?" — all reduce to that single inequality. Nothing about the data being "bad", skewed or erroneous is required: a perfectly ordinary value in the lower half of any distribution qualifies.
Three examples, three fields
- An exam. Mean 60, standard deviation 10, your score 45: z = (45 − 60) / 10 = −1.50. From the negative z score table, Φ(−1.50) = 0.0668 — the 6.68th percentile. Below average, clearly, but about 1 candidate in 15 scores lower still.
- Temperature. A January day averaging 15°C with σ = 5°C records −2°C: z = (−2 − 15) / 5 = −3.40. That is Φ(−3.40) = 0.0003 — a genuinely extreme cold day, the kind that makes news. The negative value in the raw data is a red herring; what matters is distance below the mean.
- Heights. A child measured against a growth standard at z = −2 sits at the 2.28th percentile — the conventional flag line in WHO growth monitoring. The full interpretation of magnitudes is on the negative z score meaning page.
What a negative z score is not
It is not a negative amount of anything — you cannot score "minus marks" by earning z = −1.5; you scored 45 real marks that happen to sit below the group mean. It is not a small z score: −2.5 is a large, rare score (0.62nd percentile), further from the mean than +1.96. And it is not a mistake flag: software that returns a negative z for a below-mean value is working correctly. If your z comes out negative for a value you know is above the mean, the error is in the inputs — usually μ and X swapped, or a sample mean used where the population mean belongs. The interpretation routine on this blog includes that sign check as step one.
Also asked as: can a z score be negative · can z scores be negative · can zscore be negative · what conditions would produce a negative z-score · what does a negative z score mean.
Related questions
Frequently asked questions
What percentile is a negative z score?
Always below the 50th. The mapping is the table itself: z = −1 is the 15.87th percentile, −1.645 the 5th, −1.96 the 2.5th, −2 the 2.28th. The more negative the z, the smaller the percentile.
Can a z score be negative in a hypothesis test?
Yes — it means the sample came out below the null value. A left-tailed test uses it directly (p = Φ(z)); a two-tailed test uses its absolute value. The sign carries the direction of the evidence, which is information, not an error.